What a Web Games Portal Actually Keeps Per Thousand Plays in 2026

Web game eCPMs run far below in-app rates, and portal revenue splits stack on top. Here is what an ad-funded HTML5 games portal really keeps per thousand plays. The numbers below are all sourced, and the gap between them is the whole story.

Almost every web games business plan I have seen makes the same mistake in the same cell of the same spreadsheet. Someone finds a rewarded video benchmark, sees a number in the teens or twenties, multiplies it by projected plays, and builds everything downstream from that figure. Then the first real payout arrives and it is a fraction of the model. The traffic was fine. The games were fine. The benchmark was from a different business.

📉 The Benchmark You Copied Is From a Different Business

Rewarded video in a native mobile app and rewarded video in a browser are not the same inventory, and they do not clear at the same price. AppLixir's mid-year 2026 review puts its web and HTML5 rewarded video network at a $3.62 global eCPM, $6.20 across tier-1 markets (US, UK, Canada, Australia), and $6.98 in the United States alone.

The same report cites industry mobile benchmarks of $16.49 on US Android and $19.63 on US iOS, with the broader in-app range running $15 to $40. So the US web figure is roughly a third to a quarter of the US in-app figure, for the same ad format, at the same moment in time.

That said, not everyone measures it that way, and the disagreement is large enough that you should know about it before you pick a number. Cinevva's 2026 web monetization guide, drawing on Playgama's breakdown, reports US web rewarded eCPMs of $15 to $28, EU rates of $8 to $15, and tier-3 markets like India and Brazil at $1 to $3 — explicitly as gross rates before any revenue split is applied.

Both can be true. One is a network's realised blended average across everything it serves; the other is a gross rate range before splits, skewed toward whatever the strongest placements clear. The honest conclusion is not to pick the flattering one. It is this: the plausible range for US web rewarded video in 2026 spans roughly $7 to $28 depending on whose measurement and which stage of the split you are looking at, and anyone quoting you a single confident number without saying gross or net is selling something.

Model the low end. If the business only works at the high end, it is not a business, it is a bet on ad market conditions you do not control.

🌍 Geography Sets the Ceiling Before Your Game Does

Here is the part that quietly decides whether a portal works: you cannot out-design your traffic mix. A tier-3 audience playing beautifully retained games will earn less per thousand impressions than a mediocre catalogue in front of US users. The spread in the Playgama figures — $15 to $28 in the US against $1 to $3 in India and Brazil — is not a rounding difference. It is an order of magnitude.

The Tenjin Ad Monetization Benchmark Report 2026 shows how lopsided the revenue concentration is in Q1 2026. On iOS, the United States accounted for 56% of ad revenue, with Japan at 11% (up from 9% the prior quarter), Russia 6%, the UK 5%. On Android the picture is far more fragmented: the US at 29%, Russia 13%, Japan and Brazil 5% each, and roughly 32% spread across everything else. Tenjin also puts Android at 57% of ad revenue share against iOS at 43%, drawn from a sample spanning billions of impressions and excluding kids apps.

Two practical consequences for anyone running a portal or licensing a catalogue into one:

• Traffic geography belongs in the model before catalogue size does. A 1,000-title catalogue serving tier-3 traffic can earn less in absolute terms than 150 titles serving tier-1 traffic. Volume is not the variable you think it is.
• Emerging-market portals need a second revenue line from day one. If your audience is predominantly in markets clearing low single-digit eCPMs, ad revenue alone will not carry the business. Carrier billing, subscriptions or a licensing model has to be in the plan at launch, not bolted on after the first disappointing quarter.

✂️ The Splits Stack, and They Multiply

The eCPM is the top of the funnel, not the bottom. What reaches your bank account is the eCPM after the ad network's cut, after the portal's cut, and after whatever the traffic source takes. These do not add. They multiply.

Cinevva's 2026 compilation of portal terms lays out the range of developer shares across the major web game platforms:

• Poki — a straight 50/50 split on traffic Poki sends, but developers keep 100% of revenue from players who arrive through their own channels.
• CrazyGames — 60% of in-game advertising revenue and 70% of in-game purchase revenue, per the terms published for its 2026 GameMaker web jam.
• Playgama Bridge — a stated 80% developer share.
• GameMonetize — around 45%, at the low end of the range.
• itch.io — developers set the platform's cut themselves, defaulting to 10%.

Run the arithmetic on the pessimistic path. Take the AppLixir US web figure of $6.98, apply a 50% portal split, and a thousand rewarded impressions from US players is worth about $3.50 to you. Take the optimistic path — $28 gross on the Playgama range with an 80% share — and the same thousand impressions is worth $22.40. That is a six-fold spread, and none of it is about how good your games are.

The Poki structure is the one worth studying, because it prices the thing that actually matters. Traffic you bring yourself is worth double traffic the portal brings you. Every distribution decision you make should be read through that lens: are you renting an audience, or building one?

And a sobering datapoint on the long tail from the same compilation: roughly 80% of paid itch.io games earn under $50 a month. Well-performing casual titles on the major portals land in the $200 to $2,000 per month range, while the top studios on Poki reach up to seven figures annually. The distribution is brutally uneven. Median outcomes, not top-decile outcomes, are what a portfolio model should assume.

🎯 Rewarded Video Is the Only Format Worth Fixing First

If you are going to spend engineering time on one placement, make it rewarded video. The performance data is not close.

AppLixir's H1 2026 numbers show a 97% opt-in rate, 93.8% global completion rising to 95.4% in tier-1, and a 95.1% fill rate. Adding rewarded video lifted ARPDAU by 40%, and D1/D7 retention among rewarded-engaged users was 11% higher than among users who never touched the format. Web eCPMs on that network also grew 14% between January and June 2026.

Interstitials do the opposite of all of that. They interrupt, they train players to leave, and on the open web — where the next tab is one click away and there is no installed app to return to — the cost of an annoyed player is higher than it is in a native app. On a portal the player has no sunk cost. They churn to a competitor in under a second.

Design the value exchange properly and it stops being an ad and starts being a game mechanic: an extra life, a hint, a skip, a cosmetic unlock, a doubled score. The 97% opt-in figure is not a fluke of the format. It is what happens when players are offered something they actually want at the moment they want it. Our own notes on placement and format economics live under monetization, and the same logic applies whether you built the game or licensed it.

🔐 Consent Is a Revenue Line, Not a Legal Chore

This one gets filed under compliance and treated as a cost centre, which is backwards. AppLixir reports that consented eCPMs perform at roughly twice the rate of unconsented inventory, and that privacy-driven signal loss opened a 20–40% eCPM gap in favour of consented traffic. On its network, 97% of traffic is TCF-compliant.

Read that as a pricing mechanism. A consent flow that players actually complete is worth as much to your revenue as a doubling of your fill rate, and it is far more within your control. A badly implemented consent banner — one that is confusing, that defaults to refusal, that fires before the page has rendered — is not a legal safeguard. It is a direct cut to your effective CPM, applied to every impression, forever.

The same applies to the kids-content question. Ad rates on child-directed inventory are structurally lower because behavioural targeting is off the table, which is precisely why Tenjin excludes kids apps from its eCPM sample. If your catalogue skews young, model it separately or you will overstate the whole portfolio.

💳 The Counterweight Nobody Puts in the Model

Web inventory prices lower than in-app inventory. That is the bad news, and it is real. What rarely makes it into the same spreadsheet is the other side of the ledger: the web does not charge you a platform tax on payments.

Mobidictum's March 2026 reporting tracked direct-to-consumer payment platform Appcharge crossing $1 billion in annualised DTC transaction volume, up from $700 million in January 2026 and $500 million in July 2025 — more than doubling in six months. The figure driving that scramble: an estimated $41 million a day leaking to app store fees across top publishers.

Mobidictum's companion piece on platform regulation puts numbers on the commission structures studios are trying to escape — Google Play standard billing at 20–25% plus a 5% billing fee, web purchases on iOS at around 15%, and 0% commission on alternative app store distribution in the EU. It also reports web shop capture rates of 21.31% in Türkiye against a 19.45% European average for RPG titles, meaning roughly a fifth of total game revenue now routes around the stores entirely.

For a browser-based portal, that entire fee stack simply does not apply. A subscription, a bundle sale or a direct purchase on your own domain costs you payment processing and nothing else. That structural advantage is exactly why white-label portals and subscription models keep outperforming pure ad plays in markets where eCPMs are thin — the ad revenue per player may be a quarter of app rates, but the payment economics are meaningfully better on the other side.

🚫 Five Ways Operators Break Their Own Numbers

These are the failures that show up repeatedly, and every one of them is self-inflicted.

1. Modelling web revenue with in-app benchmarks. The single most common error, and it inflates the plan by a factor of three or four before anything else goes wrong.
2. Quoting gross eCPMs internally. If your dashboard shows the rate before the portal split and the network cut, everyone in the company is working from a number nobody will ever be paid.
3. Buying catalogue size instead of buying traffic fit. A thousand titles in front of low-eCPM traffic with no acquisition plan is a licensing invoice, not a revenue stream. Match the catalogue to the audience you can actually reach.
4. Stacking interstitials to close a revenue gap. It works for about six weeks, then retention collapses and you have less inventory than you started with. Increasing ad load to fix a demand problem is the monetization equivalent of borrowing to pay interest.
5. Treating consent as a checkbox. A 2x eCPM differential is not a compliance detail. It is one of the largest single levers on the page.

🧭 A Model That Survives Month Three

Build the spreadsheet from the bottom, not the top. Start with the eCPM your actual traffic geography can clear, take the low end of the published range, subtract the portal split and the network cut, and only then multiply by realistic plays per user. If it still works, you have a business. If it only works at $28 US gross with an 80% share, you have a hypothesis that needs testing at small scale before it deserves a catalogue budget.

Then build the second revenue line immediately. Ads alone are a thin, volatile, geography-dependent income. Portals that last pair them with something structural: a subscription, carrier billing, direct sales, or licensing the catalogue onward. The web's exemption from platform commissions is the strongest argument for doing so, and it is the one advantage that does not evaporate when ad demand softens in Q1.

Forestry Games licenses HTML5 games to portal operators, telecoms and publishers who are building exactly these models, and the licensing conversation is usually more useful when it starts from your traffic mix and target economics than from a catalogue count. If you are pricing a portal build or sizing a catalogue against realistic eCPMs, that is the conversation worth having first.

One last thing worth internalising: the numbers in this post will move. Ad markets are seasonal, privacy rules keep shifting the value of a signal, and platform commission structures are being rewritten by regulators as we speak. What will not change is the discipline — know whether your number is gross or net, know what your traffic is worth by geography, and never let a benchmark from a different business set your expectations for this one.